DPDP readiness starts with a data map because each of its main duties is about specific personal data in specific places. You can't write an accurate notice, honour a withdrawal, answer an access request, erase on time, scope a breach or apply Rule 6 safeguards to data you haven't found. A data map is the record of what personal data you hold, where each copy lives, why you hold it, who owns it, which vendors receive it and how long you keep it. Build it first, and the rest of the work has something to stand on.
This post explains why each DPDP duty depends on the map, where personal data tends to hide, how to build a map in practice, and what a good row in the map contains. The core duties start on 13 May 2027.
Why DPDP data mapping comes first
The DPDP Act, 2023 and the DPDP Rules, 2025 don't use the words "data map". They don't need to. Each duty asks a question that only a map can answer. The table below sets out the main duties, where they come from, and what the map must tell you to meet them.
| Duty | Where it comes from | What the map must tell you |
|---|---|---|
| Notice | Where it comes from: Section 5; Rule 3 | What the map must tell you: An itemised description of the personal data you collect at each point, and the purpose for each item |
| Consent and withdrawal | Where it comes from: Section 6(4) and 6(6) | What the map must tell you: Which systems and processors use data on the basis of consent, so a withdrawal reaches all of them |
| Access requests | Where it comes from: Section 11(1) | What the map must tell you: A summary of one person's data and how you process it, plus every other Data Fiduciary and Data Processor you shared it with, and what you shared |
| Correction and erasure | Where it comes from: Section 12; Section 8(7) | What the map must tell you: Every copy of one person's data, including the warehouse, exports and vendors, and the identifier that finds it in each system |
| Retention | Where it comes from: Section 8(7); Rule 8(3) | What the map must tell you: The purpose and retention period for each store, and which personal data and logs must be kept for at least one year |
| Breach notice | Where it comes from: Section 8(6); Rule 7 | What the map must tell you: Which people and which data sit in the affected system, so you can describe the nature and extent of the breach and reach the right people |
| Security safeguards | Where it comes from: Section 8(5); Rule 6 | What the map must tell you: Which stores hold personal data, so encryption, access control, logging and backups cover all of them |
| Processor contracts | Where it comes from: Section 8(2); Rule 6(1)(f) | What the map must tell you: Which vendors process personal data for you and what each one receives |
| Transfers outside India | Where it comes from: Section 16; Rule 15 | What the map must tell you: Which data leaves India, to which vendor and which country |
Two of these rows deserve a closer look. Section 11(1)(b) gives a person the right to know the identities of all other Data Fiduciaries and Data Processors you shared their data with, along with a description of the data shared. If your vendor list lives in someone's memory, you can't answer that. Section 8(7)(b) also requires you to cause your Data Processor to erase the data you gave it. Deleting a row in your own database is half the job if a CRM, an email tool and a support desk each hold a copy.
Breach notice is the other one. Rule 7 asks you to tell the Data Protection Board, without delay, the nature, extent, timing and location of the breach and its likely impact, with a detailed report within 72 hours. When a system is compromised, the first question is what personal data it held and about whom. A current map answers that in minutes. Without one, the 72 hours go on finding out.
Where personal data hides

Production databases are the easy part, because someone built them on purpose. The copies that cause trouble sit elsewhere:
- SaaS tools. CRM, support desk, email and SMS platforms, payment gateway dashboards, HRMS, applicant tracking, e-signature tools and team chat. Each one holds personal data, sometimes imported once and never reviewed.
- Spreadsheets and shared drives. Exports for a campaign, a reconciliation or a board report, shared by link and kept long after the task ended.
- Free-text fields. Support tickets, call notes, chat transcripts and comment boxes, where customers paste phone numbers, addresses and ID numbers into fields nobody designed for them.
- Backups and snapshots. Database backups, disk snapshots and old buckets that still hold data you deleted from the live system.
- Logs. Application logs that capture request bodies, error messages with email addresses in them, and access logs with IP addresses. Our log retention guide covers how long to keep them and how to keep them lean.
- Analytics and SDKs. Product analytics, attribution and crash-reporting SDKs in your apps, each sending events and device identifiers to a vendor.
- Warehouses and staging. Data warehouse tables, dashboards built on them, and test environments refreshed from production.
- AI tools. Prompts and uploads to AI assistants, LLM API calls inside your product, and AI features switched on inside SaaS tools you already pay for. Our DPDP and AI tools page covers what to check.
- Vendors and partners. Agencies, collection partners, payroll providers and anyone else you send files to, by integration, SFTP or email.
None of these is unusual. They are what a normal company builds up over a few years. The map exists to make them visible.
How to build a data map in practice

A data map is built from several sources, because no single source is complete. Interviews tell you what people believe happens. Configs and scans tell you what actually happens. You need both.
1. Start with a system inventory
List every system that might hold personal data before you look inside any of them. Good places to look:
- Your single sign-on dashboard, which lists the apps people log in to.
- Finance records for software subscriptions and cloud bills, which catch tools bought on a card without IT knowing.
- Your cloud accounts: databases, storage buckets, queues, log stores and backups, region by region.
- Integrations and API keys in your code and CI, which show where data is sent.
- The SDKs listed in your mobile and web app builds.
2. Interview the people who run each system
Keep interviews short and specific. For each system, ask what personal data goes in, where it comes from, why it is needed, who can see it, where it is sent next, and when it is deleted. Talk to engineering, product, support, marketing, sales, HR and finance. The person who set up a tool may be the only one who knows what was imported into it.
3. Run discovery scans
Interviews miss things, so check the data itself. Read schemas for fields that hold personal data. Scan sample content, with approval, for patterns such as email addresses, Indian phone numbers, PAN formats and 12-digit Aadhaar-style numbers. Search a day of logs for the same patterns. Cloud providers offer scanning services that help here, such as Amazon Macie for S3 and Google Cloud's Sensitive Data Protection. Scans find the free-text fields and stray exports that nobody mentioned.
4. Draw the data flows
A list of systems isn't enough. Pick one person, such as a customer who signs up, buys and then closes their account, and follow their data from the moment it is collected to every copy, every vendor and every point where it should be deleted. Draw it. Mark where data crosses into a vendor and where it leaves India. Flow diagrams are what make erasure and breach questions answerable, because they show the path, not just the stops.
5. Reconcile what you heard with what you found
Where an interview and a config disagree, trust the config and record the difference. "We delete closed accounts after 90 days" and a warehouse table holding five years of closed accounts is a finding, not a contradiction to smooth over.
6. Keep it current
A map that was right in January is wrong by June. Tie updates to the events that change it:
- A new vendor or SaaS tool goes through procurement.
- A new SDK, integration or AI feature ships.
- A new field that holds personal data is added to a form or a table.
- A vendor changes its terms, region or sub-processors.
Give the map an owner, keep it somewhere the whole team can read, and review it on a fixed schedule. A short question in your pull request template, such as "does this change collect, store or send personal data?", catches a lot of drift.
What a good map row contains
Each row describes one system, or one dataset inside a system. The fields below are the ones the duties in the first table rely on. The example is illustrative, a made-up support desk, not a real company.
| Field | What it records | Example |
|---|---|---|
| System and dataset | What it records: The tool and the specific data in it | Example: Support desk: tickets and attachments |
| Data categories | What it records: An itemised list of personal data fields | Example: Name, email, phone, order ID, free-text messages, uploaded files |
| Data Principals | What it records: Whose data it is | Example: Customers |
| Purpose | What it records: Why you hold it | Example: Resolving customer support requests |
| Basis | What it records: Consent or the legitimate use relied on | Example: Consent at sign-up; legal review pending |
| Source | What it records: Where the data comes from | Example: Web form, email, in-app chat |
| Owner | What it records: The named person responsible | Example: Head of support |
| Location and region | What it records: Where it is stored | Example: Vendor cloud, Singapore region |
| Processors | What it records: Vendors that receive it | Example: Support desk vendor; its AI reply feature, which is on |
| Flows out | What it records: Where it goes next | Example: Nightly export to the data warehouse |
| Retention | What it records: How long it is kept and how deletion happens | Example: No limit set; manual deletion only |
| Identifier | What it records: How to find one person's data | Example: Email address or customer ID |
| Safeguards | What it records: Encryption, access and logging | Example: SSO and MFA on; audit log kept 90 days |
| Gaps | What it records: What the row shows is missing | Example: No retention period; warehouse copy not covered by erasure |
The last column is the point. Once rows are filled in, the gaps tend to show themselves: no retention period, a vendor nobody listed, a copy that erasure doesn't reach, logs that won't last a year. That list becomes the start of your gap register. Our Rule 6 checklist sets out the minimum safeguards to check each row against.
Mapping AI tools and AI vendors
AI tools are the part most inventories miss. They rarely look like a new system. An assistant is bought on a card, an AI feature is switched on inside a support desk you already pay for, or a developer adds an LLM API call to a product flow. None of these goes through the steps above unless you look for it.
Find them in the same places as other systems, plus a few more:
- Card spend and expense claims. Assistant subscriptions often start on one person's card before anyone signs a contract.
- Admin settings in SaaS tools. Support desks, CRMs, email and meeting tools now ship AI features that send content to a model. Check whether each one is on.
- API keys and outbound calls. Search your code, secrets store and egress logs for calls to model providers.
- Browser extensions and meeting recorders. Both can read pages, mail and calls, and send them to a vendor.
An AI tool needs more than the usual row. Record what goes in (prompts, uploads, transcripts, retrieved documents), whether the vendor trains on your content, how long it keeps prompts and logs, where data is stored and where it is processed, and which plan you are on. Storage and processing can differ. As of 20 November 2026, ChatGPT Enterprise offers storage at rest in India, but its inference residency options cover the EEA and Switzerland, the US and the UAE, not India. These answers change by plan and by date, so record when you checked.
Our DPDP data mapping template has a separate sheet for AI tools with these fields, linked to the system rows they sit inside. Our DPDP and AI tools page covers what to change once you have found them.
Mistakes that make a map hard to use
- Mapping only the production database. The copies in SaaS tools, warehouses, backups and logs are where erasure and breach scoping fail.
- Categories that are too broad. "Customer data" doesn't help you write an itemised notice. List the fields.
- No identifier per system. If you can't find one person in a system, you can't answer their access or erasure request there.
- Leaving out employees and candidates. Their data is personal data too, and it sits in HRMS, payroll, applicant tracking and email.
- A one-off spreadsheet. Without an owner and update triggers, the map goes stale and stops being trusted.
Where to start before 13 May 2027
If you have no map today, a sensible order is: inventory the systems this month, interview owners and scan the largest stores next, then draw the flows for your main customer journey and your employee journey. That is enough to start the work that takes longest, such as consent logs, retention jobs and one-year log retention. A longer retention setting can't bring back logs that were deleted before you changed it, so those settings can't wait for a perfect map.
Where Mono helps
Data mapping is the first part of our DPDP gap assessment. We build the personal data inventory and data-flow diagrams with read-only access, including SaaS tools, vendors and AI tools, then compare each system against the Act and Rules and give you a gap register and a costed plan. You can read what the full set of deliverables looks like in what a DPDP technical audit delivers. When the gaps are clear, we can build the fixes with DPDP implementation: consent logs, rights requests, retention jobs and breach workflows.
This article is general information, not legal advice. Our partner law firm gives the written legal opinion on how the Act applies to you.
Sources: Digital Personal Data Protection Act, 2023, Sections 5, 6, 8, 11, 12 and 16. Digital Personal Data Protection Rules, 2025, Rules 3, 6, 7, 8 and 15 (G.S.R. 846(E), Gazette of India, 13 November 2025). AWS documentation: Amazon Macie. Google Cloud documentation: Sensitive Data Protection. OpenAI Help Center: data residency and inference residency for ChatGPT. Last reviewed 20 November 2026.
FAQ
What is DPDP data mapping?
It is a record of the personal data a company holds under the DPDP Act: which systems hold it, what fields, whose data it is, why it is held, who owns it, which vendors receive it, whether it leaves India and how long it is kept. It usually comes with data-flow diagrams that show how data moves between systems and vendors.
Does the DPDP Act require a data map?
The Act and Rules don't use the term. But duties such as itemised notices, access requests that name every processor you shared data with, erasure that reaches your processors, breach reports to the Data Protection Board and Rule 6 security safeguards all depend on knowing where personal data is. In practice, a data map is how you meet them.
Where does personal data usually hide outside the main database?
In SaaS tools such as the CRM, support desk and HRMS, in spreadsheets and shared drives, in free-text fields, backups, application logs, analytics SDKs, data warehouses, staging environments, AI tools and vendor systems. These copies are where erasure and breach scoping tend to fail.
How do you build a personal data inventory?
Start with a list of systems from your SSO dashboard, software spend, cloud accounts, integrations and app SDKs. Interview the people who run each system, run discovery scans on schemas, sample data and logs, then draw data flows for one person's journey. Where interviews and configs disagree, trust the config and record the gap.
How often should a DPDP data map be updated?
Whenever something changes it: a new vendor, SDK, integration, AI feature or personal data field, or a vendor changing its terms or region. Give the map a named owner and review it on a fixed schedule as well.
